IRIS National Finalist Innovating Neural Network

Amogh from Grade 12 has recently taken part in the Initiative for Research and Innovation in STEM (IRIS) National Fair that was held virtually. The event was a national event where thousands of students from all across India participated. Only about 4% of the participating students has qualified as National Finalists, and Amogh is one of them. We reached out to Amogh to find out more about his research and why he participated in the IRIS National Fair. 

I took part in IRIS because I thought it would be a great opportunity for me to challenge myself and interact with other students who enjoy innovating and doing research – Amogh, Grade 12 

The IRIS National Fair promotes and nurtures science and scientific research among young Indian innovators. The event is organised to encourage STEM initiatives that are seeking to explore real-life challenges, and recognises outstanding projects. This fair is the national level of a larger international event, where winners get to move on to the prestigious International Science and Engineering Fair (ISEF) in the US. As a result of his project, Amogh will be attending the next fair virtually in January 2022. 

Amogh qualified for IRIS in 2018, and this was his second time in the competition. His research titled 'The Neural Layer Bypassing Network - A Novel Neural Network (AI) Architecture to Increase the Speed of Forward Propagation without Sacrificing Accuracy or CPU Load', was undertaken this summer. The research was published before being peer reviewed. Amogh explains: “This new architecture I am proposing is inspired by the standard neural network, but the network decides when to stop the predictions, even if this is in the middle of the network, making it faster than other networks.”  

Amogh became interested in AI and ML in Grade 9 after attending a seminar in his community. Since then he has been involved in research, courses and other projects to develop his understanding of the field. The project is a software with a new architecture called ‘Neural Layer Bypassing Network’, which for the time being is at a theoretical stage and will require work before becoming a commercial product or service. “Specifically for this research, I found a need for real-time AI predictions to be as fast and accurate as possible”, explains Amogh. “Unfortunately, there is a trade-off in this scenario, so I tried creating a new architecture that would minimize this trade-off, making predictions faster while still being accurate”, says Amogh, whose audience are both other researchers and companies.

Amogh is excited about being selected for the ISEF and is looking forward to carrying on with his research. “It is an honor to be selected as a national finalist at ISEF, and I will continue to pursue other endeavors with a similar rigor in the future.” To get a better understanding of what Amogh’s research is about, see the YouTube video that he has compiled.  

To find out more about the competitions and how to join, visit the IRIS website.